svdq-int4-flux.1-fill-dev is an INT4-quantized version of FLUX.1-Fill-dev, which can fill areas in existing images based on a text description. It offers approximately 4× memory savings while also running 2–3× faster than the original BF16 model.

Overview of SVDQuant. Stage1: Originally, both the activation X and weights W contain outliers, making 4-bit quantization challenging. Stage 2: We migrate the outliers from activations to weights, resulting in the updated activation and weight. While the activation becomes easier to quantize, the weight now becomes more difficult. Stage 3: SVDQuant further decomposes the weight into a low-rank component and a residual with SVD. Thus, the quantization difficulty is alleviated by the low-rank branch, which runs at 16-bit precision.

(a) Naïvely running low-rank branch with rank 32 will introduce 57% latency overhead due to extra read of 16-bit inputs in Down Projection and extra write of 16-bit outputs in Up Projection. Nunchaku optimizes this overhead with kernel fusion. (b) Down Projection and Quantize kernels use the same input, while Up Projection and 4-Bit Compute kernels share the same output. To reduce data movement overhead, we fuse the first two and the latter two kernels together.

• Developed by: MIT, NVIDIA, CMU, Princeton, UC Berkeley, SJTU and Pika Labs
• Model type: INT W4A4 model
• Model size: 6.64GB
• Model resolution: The number of pixels need to be a multiple of 65,536.
• License: Apache-2.0

Please follow the instructions in mit-han-lab/nunchaku to set up the environment. Then you can run the model with

import torch
from diffusers import FluxFillPipeline
from diffusers.utils import load_image

from nunchaku.models.transformer_flux import NunchakuFluxTransformer2dModel

image = load_image("https://huggingface.co/mit-han-lab/svdq-int4-flux.1-fill-dev/resolve/main/example.png")
mask = load_image("https://huggingface.co/mit-han-lab/svdq-int4-flux.1-fill-dev/resolve/main/mask.png")

transformer = NunchakuFluxTransformer2dModel.from_pretrained("mit-han-lab/svdq-int4-flux.1-fill-dev")
pipe = FluxFillPipeline.from_pretrained(
    "black-forest-labs/FLUX.1-Fill-dev", transformer=transformer, torch_dtype=torch.bfloat16
).to("cuda")
image = pipe(
    prompt="A wooden basket of a cat.",
    image=image,
    mask_image=mask,
    height=1024,
    width=1024,
    guidance_scale=30,
    num_inference_steps=50,
    max_sequence_length=512,
).images[0]
image.save("flux.1-fill-dev.png")

Work in progress. Stay tuned!

• The model is only runnable on NVIDIA GPUs with architectures sm_86 (Ampere: RTX 3090, A6000), sm_89 (Ada: RTX 4090), and sm_80 (A100). See this issue for more details.
• You may observe some slight differences from the BF16 models in detail.

If you find this model useful or relevant to your research, please cite

@inproceedings{
  li2024svdquant,
  title={SVDQuant: Absorbing Outliers by Low-Rank Components for 4-Bit Diffusion Models},
  author={Li*, Muyang and Lin*, Yujun and Zhang*, Zhekai and Cai, Tianle and Li, Xiuyu and Guo, Junxian and Xie, Enze and Meng, Chenlin and Zhu, Jun-Yan and Han, Song},
  booktitle={The Thirteenth International Conference on Learning Representations},
  year={2025}
}